The AI trap.

A service company I know built its own AI tool, an operations dashboard. It worked. Running the business got much easier. So they had the obvious thought: if this helps us, it would help others like us.

They asked around. A few owners of companies in the same line of service, people they already knew, said "yeah, I'd buy this." So they sold it. Six customers, all people they knew, and real money coming in.

Then they got stuck.

That's the AI trap. AI has made software cheap to start. A capable person can build in days what used to take a contractor months, and for a small company that's a gift. Use it. But the cost of software was never mostly in the first week, and the cost of selling software is a different thing again.

Think about buying a suit. Off the rack, a good one fits most people most of the way, and a tailor does the rest. Now imagine sewing machines got free and anyone could make a suit in a weekend. You'd make one cut exactly to you. A friend admires it, so you make him one. Then five more friends. Without meaning to, you've opened a tailor shop in the back of your business, and you don't have a tailor.

What it cost them

Two things happened, and they pull against each other.

First, they pulled time away from their actual business to build a business underneath it. Without quite deciding to, they became an operational software company on the side. It did make money. Others in their region with a similar business bought it.

Second, and this is the part people miss, they ended up with a less differentiated operating model. The dashboard had been their edge. It was extra gas in the tank, the efficiency that let them run better than the next company over. Then they handed it to people operating in the same market. Every sale made them a little more like their competitors.

A tool, but not a product

The other problem showed up after the sixth sale. They'd built for a very close group and nothing wider. There was no way to take it further.

Selling software to strangers takes things they didn't have. People to ship updates. Someone checking for security vulnerabilities. A roadmap. Sales, demos, someone to pick up the phone when it breaks at a customer they've never met. That whole software-sales function is its own company, and they weren't equipped for it.

So they faced a choice: build the capability to sell and manage software, or decide the money already spent was wasted. They didn't build it. The dashboard stayed a tool, made available to the six.

What they ended up with was a tool, but not a product.

You don't want to sell tools. You want to sell products, and products require productization: availability, distribution, viability over time. If you don't have those things, it is a trap.

Superpowers, or a second business

None of this argues against building. There are two very different things you can be doing when you build your own tools with AI.

The first is giving your own people superpowers. Small tools that remove friction specific to how your company works: the glue between two systems, the report nobody can get out of the off-the-shelf product, the workflow that's yours because nobody else does business quite this way. Build those. Keep them. They're your edge, and they're what this moment is for. ("AI should be a supersuit, not a severance package" makes the same case for larger companies.)

The second is starting a software company inside your company by accident. You build a tool, people outside ask for it, you sell it, and a year later some of your best people spend their days keeping software running for other businesses. Nobody decided that's what they'd be doing. Bigger organizations get their own version of this, where internal tools pile up with nobody responsible for them. "Your company is about to have a thousand apps nobody owns" is about that one.

Map it before you build it

Before you build anything, draw the work. AI process graphing is how I do it: using AI to build, visualize and structure a workflow as a network of nodes (the steps, tasks or agents) and edges (the connections between them, or the conditional logic that decides which way the work goes).

Lay the work out that way and two things get obvious. You can see whether the tool you're excited about sits on the path that matters or off to the side of it. And you can see which parts of the graph are yours, the steps that make your business different from the one down the street, and which parts look the same everywhere. The ones that are yours are the last thing you should sell.

The question to ask

When AI tempts you to expand into software, ask one question first. Is this core to the job I intended this business to do from the beginning?

If it is, build it properly. Staff it, fund the updates, plan the roadmap.

If it isn't, ask a second question. Am I willing to change my core mission? Sometimes the honest answer is yes, and plenty of good companies started as somebody's side tool.

If the answer is no, put the tool in perspective. It's great for operationalizing yourself. It may give you an edge. It isn't your secret weapon to build wealth. It may just be an attractive distractor, pulling time, funds and focus from the business that's actually meant to get you where you want to go.

Being cool is not the same thing as being valuable.